Investigation of optimal weight gain during pregnancy: A retrospective analysis of the Japanese perinatal registry database

This study aimed to determine the weight gain during pregnancy that minimizes the predicted probability of various perinatal adverse events according to the pre‐pregnancy body mass index (BMI) and make recommendations for optimal weight gain in Japan.


INTRODUCTION
In 1980, the maternal mortality rate in Japan was the worst among developed countries (OECD), and toxemia of pregnancy was the leading cause of mortality 1 ; thus, reducing maternal mortality was the highest priority in obstetrics.In 1982, Ribeiro et al. reported that systolic blood pressure was suppressed in pregnant women who experienced the so-called Dutch famine. 2 Since then, hypocaloric therapy for toxemia of pregnancy has been investigated in Japan, and strong caloric restriction has become widespread."The Nutritional Management Guidelines for Toxemia of Pregnancy" weight recommendation for the prevention of toxemia of pregnancy was formulated in 1997 and published in 1999 by the Perinatal Committee of the Japanese Society of Obstetrics and Gynecology (JSOG). 3The recommendation was for a relatively low weight gain of 7-10 kg during pregnancy for normal bodyweight (body mass index [BMI] 18-24).Following this, the Ministry of Health and Welfare (MHLW) issued a recommendation for weight gain for appropriate birthweight in 2006 4 ; however, this was also a relatively small amount of weight gain.In Japan, this recommendation to control weight gain during pregnancy has long been used.
The mean birthweight, which had decreased compared with those in the 1980s, did not increase after 2000. 5On the contrary, maternal mortality declined and fell to one of the lowest rates among developed countries. 1Science published an article titled "Staying slim during pregnancy carries a price," which suggested that the decrease in birthweight in Japan is not a matter of being lighter but may also come at a price as a lower future height. 6Furthermore, experiencing the Dutch famine in utero increased the risk of developing noncommunicable diseases later in life, giving rise to the Barker hypothesis. 7This hypothesis proposed that adverse nutrition in early life, including prenatally as measured by birthweight, increased the susceptibility to noncommunicable diseases, such as obesity, diabetes, insulin insensitivity, hypertension, and hyperlipidemia, and complications such as coronary heart disease and stroke.Several recommendations were issued for weight gain during pregnancy to achieve an appropriate birthweight.Among them, the Institute of Medicine (IOM) recommends a weight gain of 11.3-15.9kg during pregnancy for normal-weight women (BMI: 18.5-25), 8 considerably higher than the recommendation of the MHLW.With the advent of the 21st century and the decline in the maternal mortality rate in Japan, the concept of toxemia of pregnancy has changed to pregnancy-induced hypertension and then to hypertensive disorders of pregnancy. 9nsequently, extremely low-calorie therapy is no longer supported as a prevention of hypertensive disorders of pregnancy.Even with the recommendation of the MHLW, the low birthweight (LBW) rate was still high.In response to these changing times, the Committee of the Obstetrics Practice Guidelines discontinued the previous recommendations in 2018, and the Perinatal Committee of JSOG considered changing the weight gain recommendations during pregnancy.Although the IOM recommendations for appropriate birthweight have been adopted worldwide, we hesitated to employ the IOM recommendations based on BMI, which were developed mainly for Western people, to the Japanese people, who have relatively short stature.In addition, given the relatively unique circumstance in which half of all deliveries in Japan take place in clinics, the Perinatal Committee reached a consensus that not only appropriate birthweight but also goals that are compatible with prenatal health support and delivery system in Japan should be set.Furthermore, JSOG has the Japanese perinatal registry database, and we thought it would be possible to formulate new optimal weight gains.Thus, we analyzed the perinatal database and surveyed obstetricians to determine the optimal weight gain according to the prepregnancy body physique.
In recent years, the accumulation of perinatal registry databases in Japan has provided resources to examine pre-pregnancy BMI, weight gain during pregnancy, and frequency of perinatal adverse events among women giving birth in Japan and has laid the groundwork for science-based recommendations for optimal weight gain.Despite some reports on optimal weight gain using the Japanese perinatal database, [10][11][12] recommendations for optimal weight gain may differ depending on which events are given more weight.
In determining a recommendation, the unique conditions of obstetrics facilities in Japan must be considered.In Japan, approximately half of all deliveries are performed at local primary facilities, and the other half at general hospitals and tertiary facilities; however, differences were noted in their management policies, such as the availability of emergency cesarean sections and management of preterm birth.Therefore, each facility will highly likely differ in terms of which perinatal events are associated with weight gain during pregnancy and which of these events should be emphasized in the management of weight gain during pregnancy.Therefore, we also conducted a nationwide questionnaire survey to investigate the intentions of obstetricians at different facility classes using the membership list of the Japan Association of Obstetricians and Gynecologists (JAOG), in which physicians working in primary and higher facilities are evenly enrolled.

Population sample
The JSOG database on births during 2015-2017 was used.From the 719 723 deliveries included in this database, multiple births, stillbirths, and those who had missing information on either the number of fetuses, maternal age, height, pre-pregnancy BMI, pregnancy complications, gestational age at delivery, birthweight, or mode of delivery were excluded.The following women were further excluded: women who were expected to have a high risk of outcomes of interest, such as those aged <18 years and ≥45 years; had hypertension diagnosed prior to conception or before 20 weeks' gestation and pre-gestational diabetes mellitus; had previous cesarean delivery; was taking psychotropic medicines with a risk of raising blood sugar levels; had deliveries under 28 weeks and after 42 weeks of gestation; and were estimated to have a gestational weight gain rate outside the range of 0-20 kg at 40 weeks of gestation.As described below, gestational weight gain for subjects delivering at less than 40 weeks of gestation was calculated assuming gestation continued to 40 weeks of gestation.Extremely preterm births before 28 weeks were excluded from the study because they were considered to have a large margin of error of gestational weight gain and were unlikely to be due to increased gestational weight.The final population was 419 114 deliveries, consisting of 74 665 deliveries the underweight group, 303 426 from the normal-weight group, 32 336 from the Obese 1 group, and 8687 from the obese Class 2 group.The flow chart is shown in Figure 1.

Definition of variables
BMI was calculated from the pre-pregnancy weight and height and categorized into underweight (BMI < 18.5), normal weight (18.5 ≤ BMI<25), Obese 1 (25 ≤ BMI < 30), obese ≥2 (30 ≤ BMI) in line with the Japan Society for the Study of Obesity categories. 13All analyses were conducted and stratified by these four categories.
F I G U R E 1 Flow chart of the study population.The database of the Japan Society of Obstetrics and Gynecology registered 239 866 deliveries in 2015, 243 382 in 2016, and 236 475 in 2017.Of the 719 723 deliveries in total over the 3-year period, multiple births, stillbirths, and missing maternal basic information and delivery information were excluded.Of the remaining 565 667 deliveries, maternal age <18 or >45 years, certain medical history, use of certain oral medications, gestational age < 28 weeks or >42 weeks, and data missing or out of range of maternal weight gain during pregnancy, were further excluded.A total of 419 114 deliveries was the subject of this study.
The effect of interest was expected gestational weight gain at 40 weeks, calculated using a linear prediction model that estimated the "expected gestational weight gain at 40 weeks" based on the gestational weight gain at delivery and gestational length, a method previously explained. 9We conducted this calculation rather than calculating weekly GWG rate, which has been used in previous research, as it is known that while gestational weight gain increases linearly with gestational age from around 15 weeks, it is less so early in pregnancy. 14The "expected gestational weight gain at 40 weeks" was preferred as the main effect of interest because it would consider the effect of gestational age on the differences in the actual gestational weight gain and allow for the calculation of preterm delivery risk; however, we used the actual gestational weight in the sensitivity analysis.
The following nine outcomes of interest were considered: preterm delivery, delivery before 34 weeks of gestation, fetal growth (small for gestational age [SGA], appropriate for age [AGA], large for gestational age [LGA]), birthweight <2500 g (LBW), birthweight ≥4000 g (macrosomia), delivery mode (vaginal delivery, emergency cesarean delivery, and instrumental delivery), atonic hemorrhage, hypertensive disorders of pregnancy, and preeclampsia.Among these outcomes, the risks of delivery before 34 weeks of gestation, LBW, macrosomia, emergency cesarean delivery, instrumental delivery, and preeclampsia were used in constructing the weighted outcome.SGA and LGA were defined as under 10 percentile or over 90 percentile of the birthweight growth chart. 15aternal age, parity, smoking, BMI, and height were considered confounders based on previous literature.Maternal age was categorized by every 5 years: 18-19, 20-24, 25-29, 30-34, 35-39, and 40-44 years.Parity was categorized as primipara and multipara.Smoking was categorized as smoking during pregnancy, quit smoking because of pregnancy, did not smoke during pregnancy but the partner smoked, neither women nor partner smoked during pregnancy, and missing data on smoking.
Height was included as a linear term in the regression models, and a linear term of BMI was also included in addition to the categorical term used in the stratification.

Institutional questionnaire
A nationwide questionnaire survey was conducted using the membership list of the JAOG, in which physicians working in primary and higher tertiary facilities are uniformly enrolled.The questionnaire survey was conducted using Google Forms, and postcards with a QR code were sent to members to access the URL of the questionnaire.The survey asked on a scale of 1-5, which of the following nine perinatal adverse events were weighed in the weight management during pregnancy: fetal stunting and LBW, macrosomia and heavy for date (HFD), cephalopelvic disproportion, shoulder dystocia, preterm delivery (<28 weeks), instrumental delivery, emergency cesarean section, and hypertensive disorders of pregnancy.The survey questionnaire was distributed to a total of 2210 institutions.

Statistical analysis
The analysis was conducted in three steps: (1) calculation of the marginal risk of each outcome by gestational weight gain, (2) calculation of weights, and (3) calculation of the optimal weight gain.

Calculation of the marginal risk of each outcome by the gestational weight gain
For each of the nine outcomes, we constructed a restricted cubic spline model with five knots (at the 5th, 25th, 50th, 75th, and 95th) to estimate the association between the "expected gestational weight gain at 40 weeks" and outcome risks.A logistic model was used for outcomes such as preterm delivery, delivery before 34 weeks of gestation, LBW, macrosomia, atonic hemorrhage, gestational hypertension, and preeclampsia.A multinomial logistic model was used for the delivery method and fetal growth.By using these models, we generated marginal models that predicted what percentage of the population would have that outcome if they were all never smokers and had the given the "expected gestational weight gain at 40 weeks" with other characteristics unchanged.Upon generating these models, integrals of the "expected gestational weight gain at 40 weeks" were used.
As a sensitivity analysis, we repeated this analysis for all outcomes, except for preterm delivery and delivery before 34 weeks of gestation, using the "actual gestational weight gain" instead of the "expected gestational weight gain at 40 weeks" as the effect and including gestational week as a linear term in the cubic spline models.We further conducted sensitivity analyses excluding 15 995 patients (3.8% of sample) who had thyroid disease or connective tissue disorder.All analyses were stratified by the five pre-pregnancy BMI categories.

Calculation of weights
For the importance of questionnaire answers as an outcome, we calculated the overall average and the average stratified by the type of institution, that is, perinatal medical center, general hospital, and local clinic, and defined them as the weight of each outcome.
Calculation of the optimal weight gain By using the marginal risk of each outcome for each integral of the "expected gestational weight gain at 40 weeks" as calculated in (1) and the weights calculated in (2), we calculated the weighted marginal risk for each integral of the "expected gestational weight gain at 40 weeks" stratified by the pre-pregnancy BMI categories.We defined optimal weight gain as the integral of the "expected gestational weight gain at 40 weeks" that had the lowest weighted marginal risk.
Six outcomes were used for this analysis: preterm delivery under 34 weeks of gestation, LBW, macrosomia, emergency cesarean delivery, instrumental delivery, and preeclampsia.
By using these six outcomes, the weighted marginal risk was calculated as follows: The main analysis was conducted using the weight calculated as the overall average.The analysis was repeated using the averages stratified by the type of institution.

Ethical considerations
The research using the perinatal registry database and the questionnaire survey was each approved by the Ethics Committee of Juntendo University to which the corresponding author belongs (#19-154, #19-265).Due to the character of this research, informed consent is not applicable.

RESULTS
Women were on average 32.2 years old, and 57% were primipara.When categorized by pre-pregnancy BMI, 73% were normal weight, 17% were underweight, 8% were Obese 1, and 2% were Obese ≥2.For smoking, 9.6% were smokers, 3.8% were not smokers but had partners who were smokers, 48% were never smokers, and 38% had unknown smoking status.
Overall, the LBW rate was 13.2%, the preterm birth rate was 8.4%, delivery before 34 weeks occurred in 2.2%, and the macrosomia rate was 0.8%.The rate of pregnancies with atonic bleeding was 7.0%; hypertensive disorders of pregnancy, 5.2%; and preeclampsia, 2.1%.The delivery method was vaginal or planned cesarean delivery in 79% of cases, instrumental delivery in 9.1%, and emergency cesarean delivery in 11.8%.
The baseline characteristics of the study participants by pre-pregnancy BMI categories are shown in Table 1, and the pregnancy and neonatal outcomes by prepregnancy BMI categories are shown in Table 2.

Marginal risk of each outcome by the gestational weight gain
The marginal risk of each outcome with their 95% confidence intervals for each integral of the "expected gestational weight gain at 40 weeks" for the four prepregnancy BMI categories is shown in Appendix 11a-1-9d.
An increase in gestational weight gain at 40 weeks was associated with a reduced risk of LBW for all BMI categories, with risk plateauing for gestational weight gain >13 kg for the underweight and normal-weight group, >9 kg for the Obese 1 group, and >7 kg for the Obese ≥2 group.The risk of delivery before 34 weeks of gestation showed a J-shaped curve with the lowest risk observed at a gestational weight gain of 12-13.9kg for the underweight group, 11-12.9kg for the normal-weight group, and 7-7.9 kg for the Obese 1 and ≥2 groups.
An increase in gestational weight gain at 40 weeks was associated with a higher risk of macrosomia for all BMI categories, except for the absence of change in the risk for gestational weight gain <10 kg for the underweight group and <8 kg for the normal-weight group.
The risk of instrumental delivery was similar regardless of the gestational weight gain for the underweight, normal-weight, and Obese 1 groups; however, for the Obese ≥2 group with a weight gain of ≥12 kg, the risk increased with higher gestational weight gain.
The association between gestational weight gain and risk of emergency cesarean delivery was J-shaped for the underweight and normal-weight groups, with the lowest risk at 9-11.9 kg for the underweight group and 8-9.9 kg for the normal-weight group.The risk of emergency cesarean delivery plateaued for a weight gain of 0-7.9 kg for the Obese 1 group, from which it increased with higher gestational weight gains.The risk increased with a higher gestational weight gain from 0 kg for the Obese ≥2 group.
The association between gestational weight gain and risk of preeclampsia was J-shaped for the underweight Weighted marginal risk ¼ P ðmarginal risk of outcomeÞ Â ðweight of outcomeÞ f g P ðweight of outcomeÞ : and normal-weight group, with the lowest risk at 6-7.9 kg for the underweight and normal-weight groups.
The risk plateaued for weight gains of 0-7.9 kg for the Obese 1 group, from which it increased with higher gestational weight gains.The risk increased with higher gestational weight gains from 0 to 13.9 kg for the Obese ≥2 group.Sensitivity analysis estimating marginal risk based on the model using actual gestational weight gain adjusted by gestational length instead of the "expected gestational weight gain at 40 weeks" showed no large differences between estimations by the two models (results not shown).

Calculated weights
Among the 2210 facility managers to whom the survey was distributed, 616 responded.The ratio of facilities to which the respondents belonged to was consistent with the results of a facility survey conducted by the JAOG in 2018. 16The results of the intention survey are shown in Table 3. Nine outcomes were set and surveyed; however, no significant differences were found in the scores by the facility group, and the perinatal adverse event that members were most aware of in teaching weight gain to pregnant women was hypertensive disorders of pregnancy, and they were least aware of preterm delivery.Because adverse events similar to each other such as "macrosomia and HFD," "cephalopelvic disproportion" and "shoulder dystocia," and "preterm birth (overall)" and "preterm birth < 28 weeks" had similar responses, "macrosomia and HFD" and "preterm birth (overall)" were chosen as representatives and used for weighting.

Optimal weight gain
The weighted marginal risks for each integral of the "expected gestational weight gain at 40 weeks" for the four pre-pregnancy BMI categories are shown in Figure 2a-d.Actual numbers are shown in Appendix 2. The weighted marginal risk showed a J-shaped curve for all the underweight, normal-weight, and Obese 1 groups, with the risk lowest at 13-13.9, 11-11.9, and 8-8.9 kg for the underweight, normal-weight, and Obese 1 groups, respectively.For the Obese ≥2 group, the risk plateaued at ≤5 kg.The weighted marginal risk using weights based on the averages of the questionnaire answers stratified by the type of institution showed that weighting using answers from any specific institution type did not alter the results from those using the overall average (Appendix 3a-d).

DISCUSSION
We analyzed perinatal databases and surveyed obstetricians to determine the optimal weight gain according to pre-pregnancy physique.Graphs showing the risk of perinatal adverse events associated with weight gain during pregnancy by the pre-pregnancy BMI obtained in this study were reviewed by the members of the Perinatal Committee of the JSOG.From the resulting J-shaped quadratic curve, we recommended the guide for the optimal weight gain during pregnancy with 1 kg above and below from the amount of weight gain that minimizes the risk of perinatal events (Table 4).However, in the Obese ≥2 group, as the weight gain during pregnancy increases the risk of obstetric complications steadily, we gave guidance on weight gain on an individual basis and recommend up to approximately 5 kg.Compared with the previous recommendation, higher weight gain is optimal for the underweight, normal-weight, and Obese 1 groups.The increase in weight during pregnancy for each prepregnancy BMI that results in a lower incidence of perinatal adverse events in this study was largely consistent with those in previous studies using heat maps; however, the upper and lower limits of the recommendations for the underweight and normal-weight groups could be increased by an additional 2 kg. 17On the contrary, compared with the IOM recommendation, 8 the amount of weight gain for the Obese ≥2 group was smaller, and the trend was similar for other categories, as if to accommodate the short stature of the Japanese women.The prenatal health support and delivery system in Japan is characterized by numerous deliveries at clinics.While clinic deliveries are easily accessible by pregnant women, they lack sufficient medical resources to respond quickly to emergencies during delivery.Therefore, obstetricians would place more weight on emergency cesarean sections among perinatal events; however, the weighting of perinatal events did not differ significantly between obstetricians in clinics and those in perinatal centers.Therefore, the optimal weight gains developed in this study are not Note: Average of answers for "fetal growth restriction and LBW" used for the weight of "LBW," average of answers for "instrumental delivery" used for the weight of "instrumental delivery," average of answers for "emergency cesarean delivery" used for the weight of "emergency cesarean delivery," average of answers for "preterm delivery" used for the weight of "preterm delivery below 34 weeks," average of answers for "hypertensive disorders of pregnancy" used for the weight of "preeclampsia," average of answers for "macrosomia and heavy for date" used for the weight of "macrosomia."Abbreviation: LBW, low birthweight.
a Used as weight in the calculation of the weighed marginal risk.
significantly different from that of the IOM, considering that Japanese women of relatively short stature are the target population.If these optimal weight gains are recommended, the number of LBW infants is expected to reduce.An increase in the weight gain during pregnancy from the previous recommendation of 9-11 kg for the normalweight group would result in a decrease of approximately 2% of LBW (Appendix 1,1b).Conversely, a concern was that the optimal weight gain may increase the number of macrosomia (birthweight ≥4000 g).Even if primipara of normal weight gains an additional 2 kg during pregnancy, the incidence of macrosomia will increase by approximately 0.2%.The majority of institutions providing prenatal healthcare in Japan are screening for glucose metabolism abnormalities according to the International Association of Diabetes and Pregnancy Study Group recommendations and intervening in pregnant women with The lowest weighted marginal risk showed at 13-13.9 kg for the underweight group, 11-11.9kg for the normal weight group, and 8-8.9 kg for the Obese 1 group.For the Obese ≥2 group, risk plateaued at 5 kg and under.
gestational diabetes. 18This intervention has also been reported to reduce HFD 19 ; thus, this optimal weight gain is expected to be highly effective in reducing LBW infants.This study has some limitations.The optimal weight gain does not take into account the long-term prognosis of the child.However, in Japan, birthweight is related to the development of hypertension and diabetes in adulthood, and this study may indirectly demonstrate that optimal weight gain reduces hypertension in adults. 20In addition, pregnant women with overt diabetes or multiple pregnancies were excluded from this study; thus, we were unable to derive these recommendations.And this analysis follows the diagnostic criteria for pregnancy induced hypertension at the time of database entry and does not include pregnant women with hypertensive disorders in pregnancy diagnosed prior to conception or before 20 weeks' gestation.Furthermore, the recommendations were for the entire gestational age and could not lead to recommendations for weight gain at each gestational week or at each antenatal visit.However, the prognosis of pregnancy possibly differs depending on how much weight is gained at what gestational age thus resolving these issues will be is a future task.As a gestational weight gain reference chart has recently been made for Japanese pregnant women, 14 this may be possible with use of longitudinal gestational weight gain data such as is being recorded on the maternal and child handbook, and pregnancy outcome data.
In conclusion, we have formulated a weight gain during pregnancy that minimizes perinatal events according to pre-pregnancy physique.By recognizing and targeting this weight gain by pregnant women, various perinatal events will decrease, and the number of LBW neonates will decrease.Actual number of the weighted marginal risk for each integral of the the "expected gestational weight gain at 40 weeks" for the four pre-pregnancy BMI categories.

T A B L E 4
Abbreviation: BMI, body mass index; JSOG: Japan Society of Obstetrics and Gynecology.
Pregnancy and neonatal outcomes.
T A B L E 2 T A B L E 1 Baseline characteristics.Underweight; N = 74 665 Normal weight; N = 303 426 Obese 1; N = 32 336 Obese ≥2; N = 8687 T A B L E 3 Questionnaire results.